pyflink.multimodal.utils.run_image_batch_inference

run_image_batch_inference(image_series, predict_fn, mode='RGB')

Decode a native IMAGE batch, run inference, and scatter results.

mode is the model-input normalization mode. Most vision models expect RGB tensors, so the default converts valid IMAGE rows to RGB before calling predict_fn. Operators that intentionally consume another image mode should pass it explicitly.